Intel Transcript: Lip-Bu Tan on Rebuilding Silicon Valley’s Hardware Foundation, Re-architecting Memory, and Investing in Next-Generation AI Infrastructure
August 13, 2026 - Celesta Capital Tech Surge Podcast at Deep Tech Week
Introduction: Hardware is the Foundation of the AI Wave
Lip-Bu Tan: In the past, Intel missed the big wave like mobile, and then cloud AI. So from here on, I said, I am not going to miss any of the big waves.
Michael Marks: How do you feel about the fact that you are competing with these people, you are friends with them, and they are invested with you? You compete with Nvidia, and Nvidia is an investor. Now the government is an investor.
Lip-Bu Tan: And I am not the short-term guy. I am looking at 10 years, 15 years from now. You know, at my age, I can just easily retire. But I decided this is so important, I want to step in and make an impact.
Narrator: Hi everyone, this is the Tech Surge Deep Tech podcast presented by Celesta Capital. Each episode we spotlight issues and voices at the intersection of emerging technologies, company building, and venture investment. I am Michael Marks, founding managing partner at Celesta. If you enjoy Tech Surge, subscribe and leave us a review on your favorite podcast platform. Visit techsurgpodcast.com to sign up for our newsletter and find our video episodes on YouTube.
Michael Marks: Hi everyone, thanks for joining us. This episode of Tech Surge was recorded during the Deep Tech Week conference in San Francisco. Last year, the world purchased more than $790 billion worth of semiconductors. That is up more than 25% in a single year. The industry had set itself a goal of reaching a trillion dollars in annual sales by 2030. It is going to get there about four years early. For most of the time my guest has been doing this, nobody saw that coming. Lip-Bu Tan has been investing in silicon since 1987, long before this became the consensus it is today. From the outside, he funded the companies that disrupted computing. Now he is leading the one that originally defined it, and that same company might be carrying the most weight in the American technology industry. Intel is the only firm on US soil able to compete in making leading-edge semiconductors at scale, and Washington, Nvidia, and SoftBank are all now investors. Lip-Bu and I have been good friends for many years. In this chat, we talk about his investment philosophy, what he is bringing from his turnaround of Cadence Design Systems to Intel, and his priorities for the year ahead. Lip-Bu and I have known each other for, gosh, must be 25 years now.
Lip-Bu Tan: More than that, I think it is more than that.
Michael Marks: More than that. We met early on, and Lip-Bu joined the board of Flextronics, and then in 2013, we started this firm, Celesta, together with our partner, Nicholas Brathwaite. So we have known each other a long time. I just want this to be a conversation with old friends. You have done a lot of things. I think it would be helpful for the audience to hear how you got to where you are, where you started, and how you wound up in Silicon Valley. Why don't we start there?
Lip-Bu Tan’s Journey: From Nuclear Engineering to Silicon Valley
Lip-Bu Tan: Sure. I was born in Malaysia, Johor, and then spent my high school and college years in Singapore. Then I thought the energy problem was a big crisis. So I enrolled at MIT for my nuclear engineering degree. Unfortunately, the Three Mile Island accident happened, and then the placement director told me that I should get out of there, that there was no new nuclear power plant being built, and there was no future career for me. So I said okay. I was supposed to be in a PhD program, so I cut that short, and I wrote the only resume in my life to apply for a job. They sent me a free ticket and hotel to come to California, to San Francisco. I said, why not? San Francisco is a good deal. So I came here and I loved it here. I have been here for the rest of my life.
Michael Marks: How old were you when you got here?
Lip-Bu Tan: I think in my early 20s.
Michael Marks: Okay, fantastic. So you have been a technology investor and a technology pioneer for all these years. What attracted you to hardware, if you will? You have done a lot of investing around hardware from early on. What about hardware got you interested?
Lip-Bu Tan: Actually, I invest across the board. As you know, I am very curious and love to learn new things. I still remember I invested early in Google for search. I also invested in LookSmart in Australia. That first investment returned 10 times the fund, and so the rest is not as critical because when you return 10 times, every investor is happy. I always believe in building something that solves a problem the industry needs. You just zoom in, focus, find the right people to do it, and that is kind of fun, and then you figure out how to scale it. I learned a lot from Flextronics. Thank you for inviting me to be on the board.
The Contrarian Bet: Investing in Semiconductors Prior to the AI Boom
Michael Marks: Well, we have been around for a long time, and Silicon Valley started with hardware. When I listen to some of the debates, you almost forget that Silicon Valley started with Fairchild, National Semiconductor, and Intel, of course, and then went through PCs, Cisco, cell phones, and all that kind of stuff. Then, for about the last 20 years, everybody has been doing software, software, software. Of course, it takes both. Now hardware has become really important. Everybody knows Nvidia became the first trillion-dollar semiconductor company and is one of the most valuable companies in the world. That has completely changed the whole discussion around hardware. How do you think about that transition back to focusing on hardware, because obviously we have AI software, but SaaS software companies have been under stress? How do you think about that from an ecosystem standpoint?
Lip-Bu Tan: Yes, it is a very good question. First of all, I always thought that semiconductors and deep tech are very important. For a lot of applications, if you don't have the silicon or the platform, it is difficult. For example, with my phone, I had to carry three batteries because it would run out in the middle of the day unless you switched off some of the applications. In a way, I always thought that the semiconductor foundation is so important. But 20 years ago, that was not very popular. I went to some of the tier-one venture firm friends of mine, including some of yours. Initially, the whole partnership would attend the meeting with me. Then, after I talked about semiconductors, half of them would find an excuse to politely leave the room. Later on, only two would be left, sympathetically listening to me. At the end, they would ask if I had any Software-as-a-Service startup companies. It was not very popular. So I decided, being a contrarian investor—that's why I have the name Walden, because it represents being a contrarian investor—I decided to focus on semiconductors. Even some of my investors thought, are you nuts? It is a sunset industry, all the tier-one venture funds are getting out of it, and you want to double down and triple down on it? But I believed that that was the right thing to do, to be a contrarian. Low and behold, I still remember on one of the investments we made together, one of the co-investors asked me, "Name me one semiconductor company with over a $1 trillion market cap." I tried to argue with him not to shut down the semiconductor part of our business. I am so happy right now that that question is gone. The most valuable company in the world is a semiconductor company.
Learning from Boardrooms: Flextronics, Siemens, and Astera Labs
Michael Marks: Yeah, it really is. Well, one thing that's interesting—and a lot of the investors and media who are here, who are part of this whole Deep Tech Week, would agree—is there is a lot of focus, as there should be, on Nvidia, and Cerebras has gone public, and SambaNova, which was on the stage here as well. I think people tend to not think about all the little companies and all the little things that take place. For example, you talked about energy use, batteries, and stuff. There are all these little companies that we've invested in together, and that the venture guys have invested in, that make chiplets to take energy out and stuff like that. I like to remind people that the hardware and semiconductor space is a huge space. It's not just these big companies we know about. But you've been very good at picking those little companies and investing in them early. What brings you to think about it that way?
Lip-Bu Tan: Yeah, it's an interesting question. When I started, I always believed in putting in the hard work to learn what was going on, and I always have a curious mind. So, when you reached out to me to join the Flextronics board, initially I said no, but later on, I decided to do it. The main reason was really to understand the supply chain, logistics, and how to source materials. I learned a lot by being on the board and learning how to scale operations. That taught me a lot. Every time people ask me to join a board, I want to figure out what I can learn from it. For example, with Schneider Electric, I thought it was very important for me to learn why on earth Siemens bought Mentor Graphics, my competitor, and then this whole electronic design and then with Ansys, and now the software for design of the electronics and also the mechanical digital twin, and how that ties in together. Why on earth did they buy Mentor? I wanted to learn what was going on in this whole chain. It turned out to be very helpful for me in terms of power management and also industrial automation. Every company I'm involved with, I try to understand what I can learn and what I can contribute. On the startup side, looking back, we've been in the industry for a long time, and I have invested in close to 550 companies. I enjoy doing startups and finding ways to solve bottlenecks. The best thing is learning from smart entrepreneurs who want to start something, and I learn a lot by talking to them. Along the way, I accumulate experience that can hopefully be helpful to them.
Backing SambaNova and the Next Generation of AI Infrastructure
Michael Marks: The natural follow-up to that is what got you interested in SambaNova, which is now a hot company and was on stage here. Our first investment there, which was led by you, was $2 million in 2017, so the valuation was something like $12 million or some silly little number. What got you interested in that one? You have followed it along, and as I think you all know, Lip-Bu is still the chairman of SambaNova, and thank you for your work there. But what did you see back then?
Lip-Bu Tan: Yes, it is an interesting question. I was an early investor in some of the graphics companies like S3 and others, and I understood that the GPU is very power-hungry. I strongly believe agentic AI is going to be the future, much bigger than the training side. I thought that this would be an interesting problem to address. Knowing the GPU is very power-hungry—this was almost 10 years, 9 years ago—I backed two companies and two technologies that I thought would be very useful. One is Cerebras, which does wafer-scale integration. It is not easy to do, and I really admired what Andrew Feldman tried to do. I invested in the Series A. The other was SambaNova with their Reconfigurable Dataflow Unit, or RDU, architecture. It is a dataflow architecture that allows for much lower power and higher performance. I backed both of them. I liked SambaNova because the CEO and two Stanford professors came to me with the idea. In fact, I tried to persuade them. Rodrigo Liang, whom some of you know, came to my office. He used to be my customer when I ran Cadence, back when he was at Sun Microsystems. I said, "Well, we are friends, let's back you." We had backed Habana Labs, and then I said, "It's about time for you to come out and do something." He was very curious, and every time he invited me for lunch. Finally, one day he called me and said, "Lip-Bu, I'm going to start a company with these two professors with this new disruptive architecture." So, I jumped on it after doing some technical reviews, and I thought it was a very good architecture. I have been funding him from the Series A, Series B, Series C, Series D, and now the Series F round. In the Series F round, we are bringing in the right investors, closing somewhere between $800 million to $1 billion. We are delighted to see that. It is about the capital coming in, not just the valuation.
The Global Tech Ecosystem and Rebuilding Cadence Design Systems
Michael Marks: Obviously, there are themes going on here. Let's talk a bit, if we might, about the global nature of technology. You're from Malaysia, you've traveled around. There is India now, which is putting a big effort into trying to get more technology into the Indian ecosystem. We're starting to do work in Canada, as you know, which is another new area with a lot of great universities, same as we have in India. There are the geopolitical issues with China, which I'm very sad about, and I'm sure you are, because we've had such good experiences there. Everybody is trying to build their own Silicon Valleys now, and of course, there's going to be more of this because of what is happening with these companies. How do you think about the global aspects, and what impacts does that have on Intel? We'll get to Intel here in a minute. But how do you think about the global nature of our business?
Lip-Bu Tan: Yes, at Walden, when I started the firm, I always believed in the philosophy of thinking globally and acting locally. I always believe in investing where you can be hands-on and help the company. I travel a lot. I invest a lot in Israel, and I invest quite a lot in India, in companies like Mindtree and many others. It is the same thing with China and Taiwan. You always think about where the best talent is that you can get, you invest, and then you figure out how to scale the business. It is very important to just find the right team to back. In some ways, 9 out of 10 investments I make, they change their business plan along the way because the market changes. So, I back a team, not just an individual, but a team that knows how to build a world-class company with the right culture. It is wonderful to be a mentor and help them succeed. It is always great to see them go IPO or achieve successful trade sales. I still enjoy doing that.
Michael Marks: Me too. It is so much fun. We've backed companies in Israel, in Korea, and we have U-Gyu from HiDeep here in town, and of course, India, the United States, and Canada. It's exciting to see entrepreneurs from these different places because they bring different perspectives and you can learn a lot of interesting things. I agree about that. Let's switch gears a little bit. I think everybody knows what a success you had at Cadence, which is a bit different from early-stage investing. You took this company over, gosh, I think the stock was like $2.42 or some silly number, and you took it up by 3,000%. It would be interesting to hear what made that successful, and more importantly, what did you learn in that process that is helping you today at Intel?
Lip-Bu Tan: Yes, history is interesting. The former CEO and board members approached me multiple times, and they were very persistent at Cadence. Finally, I accepted to join the board, and the next thing I knew, they asked me, "Lip-Bu, can you step in to be the interim CEO?" This is a slippery slope. They said it would be for 3 months. I said no 12 times. Finally, they were very persuasive, asking me to just do it for 3 months. Those 3 months became 15 years. The stock was down to $2.42. I was doing it for 3 months as an interim CEO while conducting the search for a permanent CEO. Initially, I was part of the search committee, but I said I was too busy focusing on customers and products, so I took myself out of the search committee. They continued the search, and one day the chairman came to me with a smiling face. I said, "Okay, you found the CEO, you can pass the key to him." He said, "You can keep the key." I said, "No." Long story short, that became a 15-year journey.
Lip-Bu Tan: I managed to help turn around the culture and focus on the products. I still remember a couple of meetings with customers who were so mad. They said, "Not only do you have to return our money, but we also don't want to use your products." We won each one of them back, and they continued to expand as customers. It was my first time being a CEO. I still remember the HR director gave me three books, including "CEO for Dummies." I read every page. Looking back, not having been a CEO before actually worked to my benefit. I made a mistake in an all-hands meeting where I said, "This is my first time being a CEO, so if you have any good ideas, send them to me." I received 300 emails per day, and I responded to every single one of them. For some of them, I listened to them and had meetings with them. They said, "In the past, when we emailed the CEO, we never heard back. You surprised me by coming out to my cubicle and talking to me." I learned a lot. The culture of listening carefully to what people tell you, picking up good ideas, running with them, focusing on creating the best products, and delighting the customer is key. I still remember some customers complaining, saying, "We complain about your product and never hear back until it is time to renew the contract, and then you guys show up." I changed that. One customer told me, "Lip-Bu, I don't know what you did with them. I complained to you, and in less than 24 hours, people showed up at my office and fixed it." That is a culture of being responsive to delighting customers. That has been my trademark: being humble, being a good listener, preparing well for every meeting, taking notes on customer complaints or suggestions, and following up. That is the best way.
The Playbook for Intel: Customer Obsession and Vertical Integration
Michael Marks: And you're known for this, which is a great thing. Is it the same playbook at Intel?
Lip-Bu Tan: Yes, very much so, but it is a little bit more complex with more employees. We have the product side and we have the foundry business, so you have to drive success on both sides. Again, it is the same format: changing the culture to one of listening, being humble, learning from the customer, getting good ideas, and showing vulnerability. When you show vulnerability, people are actually willing to come and help you.
Michael Marks: For the entrepreneurs in the crowd, it seems so obvious, but so many companies don't get this right: the customer is right. Go listen to the customer, listen to what their needs are, and try to respond to that. If you do, that is how Flextronics worked, how Cadence worked, and how Intel is working. It doesn't seem complicated, but a lot of companies get this wrong.
Lip-Bu Tan: Yes, and in fact, my biggest competitor at Cadence told me, "Finally, I found out your secret." I said, "What is that?" He said, "My customers treat me as a vendor, but the same customers treat you as a partner." That is a big difference because when they view you as a partner, they share their roadmap with you, and you can also advise them and give them customer feedback. Sometime when you run a big company, you are so isolated, and I became their window. They said, "Lip-Bu, by meeting with you, I learn more about what is happening outside, how my competitors are doing, and how the industry trends are changing." That helped me a lot when I became a CEO.
Michael Marks: That's great. We have less than 10 minutes here. I would like to ask a few things about Intel, but I'm interested in strategic things, not whether the stock is going up or down. We all know what that is like. One of the things I find interesting, because of my background at Flextronics and of course with you, is that we built Flextronics on an outsourced model, and you're doing a bit more vertical integration at Intel. So, you're going to not only design and sell the devices but also make the devices. How do you think about that vertical integration, as opposed to outsourcing, because you wouldn't necessarily have to be the device manufacturer? How do you think about that vertical integration?
Lip-Bu Tan: To be candid with you, when I was approached to come and join Intel, it took me a while to comprehend that. But in the end, I decided to do it because it is an iconic company. Secondly, it is very important for the industry, and thirdly, it is very important for the United States. At my age, I could easily retire. But I decided this is so important, I want to step in and make an impact. That is the journey. Secondly, I think it is very important that the products, advanced packaging, and the foundry tie in together. You create more value for the customer when you do that, and I strongly believe that will be the case. But it is very complex. You really need to have the best products. For example, we have a very strong franchise in CPU compute. Over the years, I think we lost some of that edge. So, I invited some of the best CPU architects, GPU architects, system architects, and software engineers to come with me. You need to have a full stack, not only for the PC client but also moving into agentic AI, the edge, and physical AI, which will be the frontier going forward. You have to work tightly with all the frontier labs and AI labs so that all the software development is plugged in, allowing you to really drive efficiency for the new products you come out with. The other part is what everybody is talking about: scale, servers, data centers, and the cloud. We have a very strong presence on the server side, but over the years, we made a lot of mistakes, and now I need to correct those mistakes. We need to look at the next 5 to 10 years, see where the requirements are coming from, and figure out how to drive performance, the full-stack applications, and the layers to ensure we are really leapfrogging the next market and clusters.
Solving AI's Hard Bottlenecks: Advanced Packaging, Optics, and Cooling
Lip-Bu Tan: The previous panel talked about optical technology and high-speed connectivity, and that is where I thought, multiple years ago, you and I looked at high-speed connectivity as being very important. I backed Credo Semiconductor, which is about a $5 billion market cap today. And Astera Labs, where I backed four guys when they started from TI, is now about a $7.2 billion market cap and growing. Now photonics is becoming hot. I backed Celestial AI, and then Dust Photonics, which I just sold to Credo, and we benefited from that.
Michael Marks: It is nice when we back Credo, we back Dust Photonics, and then Credo buys Dust and they come together. This is when investing is really fun.
Lip-Bu Tan: Exactly. High-speed connectivity is becoming very important. The next thing is the CPUs and GPUs are becoming so hot, so how do you cool them? Air cooling was the standard, and then it became liquid cooling, and now microfluidic cooling. The same thing is happening with packaging. Now we are starting to look at advanced packaging and how to drive that. We have EMIB, which is our next-generation packaging, and now we are moving into glass substrates, and then artificial diamond, which is a very good thermal conductor and insulator. I am investing in that area. You just have to look at the technology, and that is the fun part of being a technology investor. Every day, I learn something new by looking at these brilliant startups from Stanford, MIT, and Berkeley. I spend a lot of time with different professors to understand what the next frontier is going to be. This is exciting for me, supporting these small companies. It helped a big company like Cadence, and now it helps Intel. We are going to be connected with the VC community and the startup community. That is how we make sure we don't miss some of the big waves. In the past, Intel missed big waves like mobile and cloud AI. From here on out, I said, I'm not going to miss any of the big waves. I want to be plugged in on that.
Michael Marks: If I understood what you said when you talked about packaging, as well as all these other extensions that you are thinking about out of the universities, is that a way of making your core business better because you get to fine-tune all these things? You look at new technologies coming out, not necessarily as discrete businesses, but as technologies that make your core business more valuable. Is that how you see it?
Lip-Bu Tan: Absolutely. You try to create a bigger platform. Intel has a very big platform for PC clients and server data, but we don't stop there. You have to think about what the future looks like. The world is changing so fast, and you have to be plugged in with all this new development and new materials. You really need to look at them, develop them in-house if you can, and if not, fund some of the startups so that eventually you can spin them in and acquire them. That will create a bigger platform.
Memory Re-architecting and the Future of Computing
Michael Marks: On that subject of what pieces you might think about bringing together into a company, I do have to go back 50 years or so, if I might, for the crowd here. You may or may not know that Intel, which was formed in 1968, was originally in the memory business, not the microprocessor business. Now, our good friend's company, SK Hynix, is worth over $100 billion today in the memory business. Do you ever give any consideration to going back into the memory business?
Lip-Bu Tan: Well, first of all, we are very excited about agentic AI and the huge demand for inference CPUs. Because of that huge demand, my days are filled with CEOs calling me saying they want to have more CPUs from us. We are truly driving the CPU business to make sure we meet that demand. Secondly, we are working on new architectures of CPUs to meet some of those requirements. Regarding CPUs and memory, there are a lot of ways we can stack them together and find new architectures for memory. In some ways, a lot of memory innovation has been stagnant, but now there are some really good areas emerging. I used to avoid investing in memory because it was a commodity business, but now it has become different. There are a lot of new technologies coming out. We are looking at new memory architectures, and it is one of my pet projects. You probably just saw the news that I hired my good friend Seok-Hee Lee, who used to run SK Hynix. So, you kind of know what I'm thinking about, even if we are not ready to unfold it yet. I always think about what the short-term, mid-term, and long-term requirements are. Usually, when I take a role in a startup company, I stay on for 10 to 12 years, and it was 15 years at Cadence. When I talk to my board, I tell them I'm not a short-term guy. I'm looking at 10 to 15 years from now. How are you going to build a bigger platform that can benefit the industry? It is fun to make some small contributions to the industry.
Michael Marks: I didn't know I was going to ask this question, but it does sound like there is a little bit of back to the future on that one.
Coopetition and Friendship in the Semiconductor Industry
Michael Marks: Well, we've got to wrap up. The last thing I want to mention and hear a little bit about is all these guys we've been friends with. Sanjay Mehrotra was at SanDisk, where I was chairman of the board, and now he is running Micron. We knew Jensen Huang when he had 30 engineers in Santa Clara. And of course, Elon Musk... and you used to be the CEO of Tesla, right?
Lip-Bu Tan: Right, that was for a nanosecond.
Michael Marks: But here we are, and we're all friends and competitors, and it is just this big mix. I get this question sometimes from the media: how do you feel about the fact that you're competing with these people, yet you're friends with them and they are invested with you? You compete with Nvidia, and Nvidia is an investor, and now the government is an investor. Isn't it just a wonderful thing?
Lip-Bu Tan: Michael, we have been friends for 40 to 50 years now, and I always value friendship. We stay friends even though we are partners in multiple areas. Sanjay Mehrotra is a very close friend of mine, along with his wife and our families. All in all, I don't view them as competitors. I view the market as big enough for all of us. Of course, we just find ways to create a bigger market for all of us so we can work with friends. If you remember that research, I am considered one of the most connected people, especially on the technology side. In some ways, it is always good to learn from friends, and some of those friends have become very good mentors to me. It has turned out to be very helpful in terms of my career development. Likewise, I am trying to spend my time with some of the younger CEOs, and I really enjoy working with them.
Michael Marks: Well, I know there are a lot of demands on your time, and you have to deal with famous politicians and go on CNBC every other day. I can't possibly thank you enough for carving out some time to do this, my friend. This has been a wonderful experience for all of us. Lip-Bu Tan, thanks, Lip-Bu.
Michael Marks: Lip-Bu has been ahead of the pack for nearly 40 years. He has always believed silicon is the foundation, and everything else runs on top of it. For the past 20 years, software got the valuations and the capital, but he kept writing checks into chip companies anyway. For long stretches, he was doing it nearly alone, and he never changed his mind. It is the rest of the industry that has finally changed theirs. American chip companies took 53% of the global market last year, the highest US share since 1984. The man who spent years explaining to skeptical rooms why semiconductors mattered is now running the most consequential semiconductor company in America, with Washington, Nvidia, and SoftBank all on the cap sheet. Being right early is lonely work. It is a lot less lonely now.
Narrator: Thanks for tuning in to the Tech Surge podcast from Celesta Capital. If you enjoyed this episode, feel free to share it, subscribe, or leave us a review on your favorite podcast platform. We'll be back every two weeks with more insights and discussions on all things Deep Tech.